Motional averaging of proton nuclear Overhauser effects in proteins

Carstens, Christian Renner, Alexander G. Milbradt, Luis Moroder, and Paul Tavan ... Lorna J. Smith , Alan E. Mark , Christopher M. Dobson , and Wi...
0 downloads 0 Views 955KB Size
J. A m . Chem. S o t . 1984, 106, 1923-1930

1923

Motional Averaging of Proton Nuclear Overhauser Effects in Proteins. Predictions from a Molecular Dynamics Simulation of Lysozyme? E. T. Olejniczak,zl C. M. Dobson,*+"M. Karplus,*f and R. M. Levye Contribution f r o m the Department of Chemistry, Harvard University, Cambridge, Massachusetts 021 38, and Department of Chemistry, Rutgers, The State University of New Jersey, New Brunswick, New Jersey 08903. Received July 5 , 1983

Abstract: A molecular dynamics simulation has been used to explore the effects of fast internal motions on the IH NMR relaxation behavior of lysozyme. It is shown that cross-relaxation rates, which are experimentally accessible from the time development of nuclear Overhauser effects (NOEs), are reduced by motional averaging on a picosecond time scale. Factors influencing the magnitude of these changes have been explored, including the degree of correlation of side-chain torsion angle fluctuations. More general effects on NOEs were examined by multispin simulation techniques. The internal motions altered the magnitude of specific NOEs by as much as a factor of 2. However, at the present level of experimental knowledge, it is difficult to separate these effects from the results expected for an internally rigid structure with a shorter correlation time for overall tumbling. The results show that in general, the picosecond internal motions examined here have little effect on estimates of distances obtained from NOE measurements.

A variety of experimental and theoretical techniques are now being applied to the study of internal motions of proteins.',* Nuclear magnetic resonance (NMK) measurements are particularly useful because they can provide specific information about both the amplitudes and time scales of these motion^.^ Relaxation times and nuclear Overhauser effects (NOEs) for 13Care of special interest because their relaxation is dominated by the fluctuating dipolar interactions between the I3C nucleus and the directly bonded p r ~ t o n . The ~ relaxation parameters can be related to the motional averaging of the dipolar interaction and hence to the motional behavior of individual group^.^^^ The available studies have demonstrated that large-amplitude fluctuations can occur on nanosecond or subnanosecond time scales. Furthermore, comparisons with the results of molecular dynamics simulations have shown that the averaging due to picosecond fluctuations can have a significant effect on the relaxation times of both protonated and nonprotonated More recently, 'H NOEs have been used to obtain motional information for proteins.1° Averaging effects due to atomic fluctuations decrease the magnitude of the cross-relaxation rates in the same way that they increase the 13C relaxation times. Molecular dynamics trajectories of proteins permit the direct calculation of the time correlation functions required for determining the effects of picosecond motional averaging. This approach has proven to be useful in I3C N M R for examining motional models used in the analysis of data and as an aid in the interpretation of A similar study of the effects of internal motions on 'H NOEs is highly desirable. Such a study could test the methods used to interpret the N M R measurements. In particular, it could examine the effect that motional averaging has on structural information obtained from N O E data. Of considerable importance is the effect that hindered motions of residues in the interior of the protein have on the 'H N O E data. Because of excluded volume effects, any large-magnitude motions that occur are a consequence of collective effects involving a number of neighboring Most models used to interpret relaxation data treat each degree of freedom as independent. This assumption is not valid in the protein interior," although it is appropriate for the behavior of exterior side chains.* 'Supported in part by grants from the National Science Foundation and the National Institutes of Health. Harvard University. Rutgers, The State University of New Jersey. Present address: Massachusetts Institute of Technology, National Magnet Lab, Cambridge, MA 02138. I' Present address: Inorganic Chemistry Laboratory, Oxford University, Oxford 0 8 1 3QR, England.

*

0002-7863/84/1506-1923$01.50/0

The present study is limited to the effect of subnanosecond fluctuations on the N O E values. It does not consider the contributions of internal motions with longer time scales because of the limited length of the available dynamics runs. Fluctuations on slower time scales will further decrease the magnitude of the relaxation rates for the spectrometer frequencies and tumbling times under consideration. Many of the effects that internal motions can have on NOEs may, however, be demonstrated by considering only the effects of the fast picosecond motions. We have focused our study on several residues found in the hydrophobic box region of lysozyme for which there are assigned resonances and detailed experimental evaluations of N O E S . ' ~ , ' This ~ region is in the interior of the protein, and the motions are expected to be governed by van der Waals interactions, which are well modeled by the empirical potentials used in the dynamics simulations. The following section outlines the theory for the relaxation rates which govern the NOEs in the presence of fast internal motions as well as overall protein tumbling. Simplifications in the theory are shown to be possible when the time scale of the internal motion is much faster than the tumbling time. We then compare the NOE results from a static structure to those obtained with the molecular dynamics trajectories and outline the conse(1) M. Karplus and J. A. McCammon, CRC Crit. Rev. Biochm., 9, 293 (1981). ( 2 ) F. R. N. Gurd and M. Rothgeb, Adu. Protein Chem., 33, 74 (1979). (3) R. E. London in "Magnetic Resonance in Biology", Vol. 1, J. S. Cohen, Ed., Wiley, New York, 1980, p 1. (4) A. A. Allerhand, D. Doddrell, and R. Komorski, J . Chem. Phys., 55, 189 (1971). ( 5 ) G. C. Levy, D. E. Nelson, R. Schwartz, and J . Hochmann, J . Am. Chem. Soc., 100, 410 (1978). (6) R. J . Wittebort, M. Rothgeb, A. Szabo, and. F. R. N. Gurd, Proc. Natl. Acad. Sci. U.S.A. 76, 1059 (1979). (7) R. M. Levy, C. M. Dobson, and M. Karplus, Biophys. J., 39, 107 ('1 982). ( 8 ) R. M. Levy, M. Karplus, and P. G. Wolynes, J . Am. Chem. SOC.,103, 5998 (1981). (9) R. M. Levy, M. Karplus, and J. A. McCammon, J . Am. Chem. SOC., 103. 9 x. 1,).. .-, 994 . . . (,1 .. (IO) E. T. Olejniczak, F. M. Poulsen, and C. M. Dobson, J . Am. Chem. Sor., 103, 6574 (1981). (11) W . F . Gunsteren and M. Karplus, Biochemistry, 21, 2259 (1982). (12) S. Swaminathan, T . Ichiye, W. van Gunsteren, and M. Karplus, Biochemistry, 21, 5230 (1982). (13) F. M. Poulsen, J. C. Hoch, and C. M. Dobson, Biochemistry, 19, 2597 (1980). (14) J. H. Noggle and R. E. Shirmer, 'The Nuclear Overhauser Effect", Academic Press, New York, 1971. ( I 5 ) A. Abragam, "The Principles of Nuclear Magnetism", Clarendon, Press, Oxford, 1961.

0 1984 American Chemical Society

1924 J. Am. Chem. SOC.,Vol. 106, No. 7, 1984

Olejniczak et al.

quences of internal motions for the interpretation of NOE measurements. Finally, the conclusions of the present study are summarized.

Theory (a) Motional Averaging of Relaxation Rates. In this section, we present the theory that describes the effect of internal motions on relaxation rates. We then outline the use of molecular dynamics results for calculating picosecond averaging effects. Finally we consider a model that treats the fluctuations about each side-chain bond as independent; this product approximation is compared to the results from the full dynamics trajectory. For a system of dipolar-coupled spins with no cross-correlation, the spin-lattice relaxation is governed by coupled differential equations of the f ~ r m ' ~ J ~

where Z z ( t ) j and lo,are the z components of the magnetization of nucleus j , pi is the direct-relaxation rate of proton i, and uij is the cross-relaxation rate between protons i and j . For dipolar relaxation pi and uij can be defined in terms of spectral densities as

molecular frame and Omol(t) and +rnol(t) are the spherical polar angles relating the interproton vector to the fixed molecular frame. For an isotropically tumbling molecule, the time correlation function has been shown to decay as a single exponential of the formI8

where 7o is the correlation time for molecular tumbling. Introducing this result into eq 5 gives

where (A(O)A(t))is referred to as the internal motion correlation function. For a rigid molecule undergoing only isotropic tumbling, the spherical harmonics on the right-hand side of eq 7 can be summed and substituted into eq 4 to yield

and uij =

6i~ -y4h2[2Jij(2w) - (l/3)Jij(0)]

(3)

5

where the subscripts ij refer to the pairwise interaction between protons i and j , y is the proton gyromagnetic ratio, and w is the 'H Larmor frequency. The spectral density functions, Jij(w), can be expressed as Fourier cosine transforms of the correlation functionsI6 Pij(W)

=

where P,(o(t)+(t)) are second-order spherical harmonics and the angular brackets represent a correlation function. The quantities olab(t) and +lab(t) are the polar angles at time t of the internuclear vector between protons i and j with respect to the external magnetic field and r,, is the interproton distance. It is clear from eq 3 that u,,, the pairwise cross-relaxation rate, is the quantity that can be interpreted most easily to obtain structural and motional information because only one proton-proton interaction is involved; p, as given in eq 2 includes a sum over a number of such interactions.'0 If the internal motions are uncorrelated with the overall molecular tumbling, the time correlation function appearing in eq 4 can be separated into contributions from molecular tumbling and internal motions

-

p m ~ o l ~ b ~ ~ ~ ' # ' l a ~ ~ ~ ~ ~ p * m ~ o l a b ~ ~ ~ ~ l ~ b ~ o ~ ~ riJ3(o)ri~3(r)

p a ( o m o ~ ( t ) + m o ~ ( t ) )p*a,(8rno1(O)+mo1(0))

ri; (O)r,3(t )

)

(5)

The Pma are Wigner rotation matrix e1ements.l' The Q l ( t )are the Euler angles that transform from the laboratory frame to the (16) I. Solomon, Phys. Reu., 99, 559 (1955). (17) D. M. Brink and G. R. Satchler, 'Angular Momentum", Oxford University Press, Oxford, 1971.

If internal motion is occurring, however, the ensemble average over molecular conformations on the right-hand side of eq 7 needs to be evaluated. (b) Molecular Dynamics Simulations and Motional Averaging. The molecular dynamics trajectory used in this investigation was obtained from a 33-ps simulation of lysozyme. The method employed in the simulations has been described previo~sly.'~Protons were not included explicitly but were taken into account by the extended-atom technique in which the mass and van der Waals radii of the directly bonded heavy atoms are adjusted to compensate for the protons." The average temperature during the dynamics run was 304 K. Details of the simulation and its analysis will be reported elsewhere.*O Three hundred and thirty coordinate sets separated by 0.1 ps were selected from the trajectory and used in the analysis. For each configuration, protons were generated in standard Methyl configurations using a C-H bond length of 1.08 group protons were constructed in a staggered configuration. To obtain a static structure with which the dynamics results could be compared, protons were generated onto the average heavy-atom coordinates of the dynamics run. The average coordinates of the heavy atoms were energy minimized for 100 steps to remove unreasonable internal coordinate valueslg before being used to generate the proton coordinates. The root-mean-square difference between the energy-minimized and average heavy-atom structure was 0.149 A. Relaxation rates and distances obtained from the average structure, assumed rigid, were compared with the averages and correlation functions computed from the 330 coordinate sets of the trajectory. The motions modeled by the trajectory are dominated by high-frequency fluctuations.I2 These fast motions were nearly homogeneous for the dynamics run for the residues studied here; Le., a comparison of interproton distances and dihedral fluctuations for the two halves of the run (the first 16.5 ps and the second 16.5 ps) gave very similar results. It is therefore expected that the results were not significantly influenced by any statistically rare events.

a.21

(18) D. Wallach, J . Chem. Phys., 47, 5258 (1967). (19) B. R. Brooks, R. E. Bruccoleri, B. D. Olafson, D. J. States, S. Swaminathan, and M. Karplus, J . Comput. Chem., 4, 187 (1983). (20) T. Ichiye, B. Olafson, and M . Karplus, to be submitted for publication. (21) J. C. Hoch, C . M. Dobson, and M. Karplus, Biochemistry, 21, 1 1 18 (1982).

Motional Averaging of NOES in Proteins

J . Am. Chem. SOC.,Vol. 106, No. 7, 1984

1925

to the spectral density function is negligible. This can be seen by approximating both cos ( u t ) and exp(-r/~,) by 1 for ( t < 7p < 7 0 ) in eq 10. The first integral is then approximately L r p [ ( A ( t ) A ( 0) )S 2 ( r i y 6 ) ]dt v

0.0

2.0

4.0

6.0

(1 1)

For an exponential decrease in the internal motion correlation functions found in this and other studies using dynamics trajectories of proteins, the integral is less than r , [ ( A ( O ) A ( O ) )S 2 ( r i j 4 ) ] / 2which , is the area of the triangle of base 7p and height [ ( A ( O ) A ( O )) S 2 ( r j y 6 ) ] . For 7, < 33 ps, 7 0 = 10 ns, and a spectrometer frequency of 500 MHz, the area of this triangle is less than 2% of the second term in eq 10 for all of the cases studied in this work. Thus this contribution to the spectral density function has been neglected here. The Fourier transform of the second integral in eq 10 gives a spectral density function that is scaled ; is, by the factor S Z ( r t 6 ) that

8.0

Ttme ( p s l

Figure 1. Decay of the fixed-distance interproton internal motion correlation function for intraresidue vectors on the four hydrophobic box residues studied (eq 7): (a) Trp-28 H0-Hq3, (b) Trp-108 Ha-Hq3, (c) Met-105 H”-H’*,

(d) Ile-98 H’11-H712.

The results of molecular dynamics simulations permit direct evaluation of the internal motion correlation function.” The decay of several internal motion correlation functions obtained from the dynamics trajectory is shown in Figure 1. Because the motions in the interior of a protein tend to be restricted by van der Waals interactions, the correlation functions generally do not decay to zero during the dynamics trajectory.’~~Instead, after a rapid initial decay a plateau value is reached in 7, picoseconds, where 7p is a time short in comparison with the length of the trajectory, so that 7 p 7p, the cosine Fourier transform of eq 7 can be evaluated by breaking the integral into two parts; we have 4ir

( A ( O ) A ( f ) e) x p ( - t / ~ ~cos ) ( u t ) dt =

L 4irI r p [ ( A ( 0 ) A ( t ) )- S2(rij”)] exp(-t/7,)

cos (ut) dt

& L m S 2 ( r i j 4 )exp(-r/.ro) cos

( u t ) dt

Pd(e3(t)@,(t))

p*&(e,(o)@,(o))

rij3(O)rjj3(t ) where Q l relates the molecule-fixed axis to the first local axis system which has its z axis along N - P , Q 2 relates the first axis to the second with its z axis along C*-CO, and Q3 relates the second axis to the third with its z axis along CO-CT; additional product terms would be introduced as the transformation moves out along the side chain of the residue being studied. The angles 03(t)and 4 1 ~ ( r ) are the polar angles of the interproton vector in the final axis system. If the motions are uncorrelated, the averages about each bond can be taken separately to give for the right-hand side of eq 13

+ (10)

where S2(riY6)was substituted for (A(O)A(r))when r > In the present cases, 7, is ony a few picoseconds and the plateau value of the correlation function, S2(ri,”), is always much greater than zero. Under these conditions the contribution of the first integral (22) G. Lipari and A. Szabo, J . Am. Chem. Soc., 104, 4546 (1982).

(23) G. Lipari and A. Szabo, J . Am. Chem. Soc., 104, 4559 (1982).

Olejniczak et ai.

1926 J . Am. Chem. SOC.,Vol. 106, No. 7 , 1984 lim ( A ( t ) B ( O ) )= ( A ) ( B )

(01

t-m

HE?

the long-time behavior of eq 14 can be rewritten as8

The cosine transform of eq 15 gives the spectral density function at long times, assuming uncorrelated motion about the bonds. (d) Multispin Simulations of the NOE. Selective saturation of a transition in a dipolar-coupled spin system results in intensity changes in the other transitions. The intensity changes are called NOEs.14 Time-dependent NOEs can be observed, for example, by measuring the change in intensity of the resonances as a function of the length of time that the saturating radio-frequency field is present. The change in the resonance intensity of spin i caused by irradiation o f j can be used to define a time-dependent nuclear Overhauser enhancement factor

(b) HYi2H3CYZ H a

(C)

HYl H 3 C'-S-Cy-C

(rz(t)

oi(t) =

- 10)i

In a large molecule like a protein the observed qi(t) will depend not only on the cross-relaxation rate uij coupling i to the irradiated proton j (eq 3) but also on the relaxation rates of the other spins in the system. It is therefore necessary to consider multispin effects in order to interpret the observed NOEs. Such multispin treatments have been described earlier24,25and are used here to simulate the NOEs. Each multispin simulation involved a partial set of the protons in the protein. The first 20 protons or methyl groups included in a set were those with the shortest distances to the irradiated proton in the proton coordinate set derived from the average coordinates of the dynamics run. To minimize the effect of a limited proton set on the calculated NOEs we also included the three nearest neighbors of each of the first 20 protons in the set. The p values (eq 2) for these additional protons included contributions from their three nearest neighbors even if their neighbors were not explicitly included in the proton set. Each resulting proton set included at least all of the protons within 4.5 8, of the irradiated proton. This approach assumes that no significant amount of magnetization which leaves the proton set will return through cross-relaxation effects. To test the assumption, proton sets of various sizes were included in the N O E simulations. It was found that proton sets that are larger than those described above did not cause significant differences in the calculated NOEs of the 20 closest protons. A complication in the analysis is the treatment of methyl groups. Here we make use of the fact that the correlation time for overall ~ 10 ns)26,27 is much slower than the correlation times tumbling ( T = for methyl rotation, which are in the range of 20-200 ps.28 In the NOE simulations it was assumed that the methyl rotation can be modeled by fast random jumps between three methyl proton sites. The spectral density function obtained between a methyl proton and a non-methyl proton for an otherwise rigid protein tumbling in solution is then given by29

(24) C. M. Dobson, E. T. Olejniczak, F. M. Poulsen, and R. G. Ratcliffe, J . Magn. Reson., 48, 97 (1982). (25) A. A. Bothner-By and J. H. Noggle, J . Am. Chem. SOC.,101, 5152 (1979). (26) S. B. Dubin, N. A. Clark, and G. B. Benedek, J . Chem. Phys., 54, 5158 (1971). (27) K. Dill and A. Allerhand, J . A m . Chem. Soc., 101, 4376 (1979). (28) E. R. Andrew, W. S. Hinshaw, and M. G. Hutchins, J . Magn. Reson. 15, 196 (1974).

HPI

I Ip-C I 1 HY2 H P 2

Ha

la